Django Q过滤器否定不会产生预期的查询

Posted 2023-02-23

【中文标题】Django Q过滤器否定不会产生预期的查询

【英文标题】：Django Q filter negating doesn't produce expected query

【发布时间】：2015-04-30 08:40:22

【问题描述】：

简化设置

class System(models.Model):
    fields...
    vulnerabilities = models.ManyToManyField('Vulnerability', through='SystemVulnerability')

class Vulnerablity(models.Model):
    name = ...
    risk = CharField(choices=['H','M','L','I'])
    ...

class SystemVulnerability(models.Model):
    vulnerability = ForeignKey
    system = ForeignKey
    fields...

sv_list = [199026, 199036, 199046, 199048, ....]
other_filter = Q(... lots of stuff...) #not including this for brevity

System.objects.filter(other_filter & Q(systemvulnerability__in=sv_list)).query.__str__()
u'SELECT "system"."id", "system"."ip", "system"."org_id", "system"."dnsname", "system"."ipownercode", "system"."securityplan" 
    FROM "system" 
    INNER JOIN "systemvulnerability" ON ( "system"."id" = "systemvulnerability"."system_id" ) 
    WHERE "systemvulnerability"."id" IN (199026, 199036, 199046, 199048, ....)'

看起来不错。如果我们否定它呢？

System.objects.filter(other_filter & ~Q(systemvulnerability__in=sv_list)).query.__str__()
u'SELECT "system"."id", "system"."ip", "system"."org_id", "system"."dnsname", "system"."ipownercode", "system"."securityplan" 
    FROM "system" 
    WHERE NOT ("system"."id" IN (SELECT U1."system_id" FROM "systemvulnerability" U1 WHERE U1."id" IN (199026, 199036, 199046, 199048, ....)))'

什么？ 我希望得到的查询是：

u'SELECT "system"."id", "system"."ip", "system"."org_id", "system"."dnsname", "system"."ipownercode", "system"."securityplan" 
    FROM "system" 
    INNER JOIN "systemvulnerability" ON ( "system"."id" = "systemvulnerability"."system_id" ) 
    WHERE "systemvulnerability"."id" NOT IN (199026, 199036, 199046, 199048, ....)'

如何修改我的语句以使结果查询更符合我的预期？

Example data

System:
s1
s2
s3

Vulnerability:
v1
v2
v3

SystemVulnerability:
sv1 = s1, v1
sv2 = s1, v2
sv3 = s2, v2
sv4 = s3, v2
sv5 = s3, v3

System.filter(Q(systemvulnerability__in=[sv1, sv4]))
Get back s1, s3: Good!

System.filter(~Q(systemvulnerability__in=[sv1, sv2]))
Get back: s2, s3: Good!

System.filter(~Q(systemvulnerability__in=[sv1, sv4, sv5]))
Get back s2: Bad!
But I wanted s1, s2

显然我可以根据 SystemVulnerability 进行查询，但我的过滤器实际上比这要复杂得多，所以我想保持它基于 System.Vulnerability。

我想System.objects.filter(pk__in=SystemVulnerability.filter(~Q(systemvulnerability__in=sv_list)).values_list('system')) 可以工作，但在 SQL 中生成的 IN 子句相当大而且速度很慢，尤其是随着我的数据增长。

编辑：我去尝试用前一行的 IN 子句来做。它有效，但最终否定了我的许多其他过滤器。我真的只需要 ~Q() 才能正常工作！

【问题讨论】：

System.objects.filter(other_filter).exclude(systemvulnerability__in=sv_list)?

同~Q() AND NOT ("system"."id" IN

【参考方案1】：

这不是一个真正的答案，但几天后我设法考虑如何在不排除的情况下重写它。我的 SQL 开发人员对我的所作所为感到痛苦，但它在 django 中的效果要比我开始做的 .extra(where=[]) 好得多。那样的话我会重写太多查询。